A (very) brief & non-technical introduction to Machine Learning

We hear Machine Learning used as a buzzword more and more frequently, but what actually is it?

Below we will endeavour to leave you feeling ready to explain what the essence of Machine Learning is to any friend or foe with confidence (or at least fake confidence) without talking anyone's head off.

Let’s start with a basic definition:

“Machine learning is a branch of artificial intelligence (AI) and computer science which focuses on the use of data and algorithms to imitate the way that humans learn, gradually improving its accuracy.” -source

Some easy examples of machine learning that most of us experience on a daily basis are:

• Recommendations of which movies to watch next on Netflix — the platform learns more and more about the type of things we like to watch so it can provide us with better suggestions.

• Another is intelligent search engines like Google — that learn overtime which are the best results to show at the top for certain key word/phrase searches.

Here are a couple of fun analogies borrowed from this article here by Martin Willers:

1.

A Fancy thing labeller machine.

Humans help teach the machine how to label things accurately by showing them lots of different images and telling them what they are, so that it can eventually do it by itself. The key here is to make sure that you are using the right examples to teach the model.

2.

An island of drunk people.

Now, this ‘Fancy thing labeler machine’ needs a lot of (the right type) of testing.

Picture one of your friends and imagine that they bought an island. Now, imagine that all of your other friends retired to this island, and now spend their days drunk on cocktails, while using their laptops that are all connected. (This is sounding like some weird dream, but bare with me).

Your friends now have lots of free time and are eager to help you, but it’s hard to get them to understand what you want them to do — and they only respond well to examples.

Before you get all these drunk friends on this island to do any work you need to consider: How drunk are these people? Can they even do this specific task?

You can’t just place all your trust in them to do the work, so you need to prepare some key things first:

• What does doing the task correctly look like?
• How do you rank mistakes on a scale of ‘absolute disaster to ‘not so bad’?
• How do you score the 1000 units of drunk replies (some imperfect) as a pile of work?

This illustrates a little bit of what it’s like to train a Machine Learning model.

Hopefully at this point you have a bit more of an overall understanding of what machine learning means, and an idea of how it works through these analogies.

If you are someone who is more technical, or you just want to get into more of the nitty gritty behind the scenes of how ML actually works — I suggest you follow our Head of Engineering, Bin Wang here on Medium! He uses lots of diagrams and visual aids to make concepts easy to grasp. There’s also this awesome article by Berkeley here too covering more of the basics.

👀 Keep an eye out for our next few blog posts where we will talk about how to get started with trying out Machine Learning (it’s easier than you think!), identifying opportunities for using machine learning, and finally a breakdown of the different stages of implementation.